girmaneshir/Topic-Amharic-Texts

This is a repo for Amharic Topic Detection # Amharic Topic Detection This is a Python program that does Topic Detection of Amharic Texts. The program consists of different components and three experimental settings. And three feature sets (LDA, TF-IDF and combinations of LDA and TFIDF) are used in those experiments. (1) preprocessing (cleaning data, tokenize, numeral removal, normalization, stop word removal) (2) Exp. I (WithNo Stemming) (3) Exp. II (With Stemming) (4) Exp. III (With SMOTE) ## Requirments Install anaconda and create vritual environment Make sure installing Python 3 and above`from anaconda Install necessary library files for machine learning : scikit learn, pandas, imblearn, etc. ## How to test (1) Download and extract zip files which contains (jupyter notebook, datasets, support files) (2) Download the data set from ""doi:10.5281/zenodo.5504175"" (3) Change anaconda virtual env. and open "jupyter notebook" from the command line (4) Upload jupyter notebook file named"Amharic_Topic_Modeling_for_Amharic_User_Generated_Texts_SEPT2021_MDPI.ipynb" from your browser of jupyter local server (5) Change the directory of the support code .py files such as "Amharic_stemmer_Tilahun_API.py" and "amharic_normalization_api.py" and data set(doi:10.5281/zenodo.5504175) of jupter notebook code by the directories of the coresponding files (in step 1) on which they are located on your computer (6) Enjoy testing it! ## How to cite You can cite this source "Neshir, Girma. Corpus for Amharic Topic Classification. 2021, doi:10.5281/zenodo.5504175."